This website exposes public, read-only tools over its rendered page index. No login is required. These tools do not connect to a WordPress installation and cannot change anything.
REST operations
| Tool | Request | Bounds |
|---|---|---|
search_site | GET /api/agent/search?q=sitemap&limit=10 | Query 1–120 characters; 1–20 results |
list_site_pages | GET /api/agent/pages?limit=25 | 1–100 results; optional pagination cursor |
get_site_page | GET /api/agent/page?path=/docs/sitemaps/ | Exact indexed pathname only; maximum 512 characters |
get_site_summary | GET /api/agent/site | Identity, description, and page counts |
curl 'https://cybermaps.dev/api/agent/search?q=static&limit=3'
curl 'https://cybermaps.dev/api/agent/page?path=/docs/sitemaps/'Invalid input returns HTTP 400, unknown indexed pages return 404, and unavailable index data returns 503. The API uses a Cloudflare rate-limit binding with 60 requests per minute per IP at an edge location. Limits are approximate and do not form a global quota. A 429 response includes Retry-After.
MCP
The same four tools are available at /mcp using the SDK’s Streamable HTTP transport and protocol version 2026-07-28. Use server/discover and tools/list for live capabilities. Tools are read-only, non-destructive, idempotent, and closed-world. The interface is stateless and does not expose subscriptions.
Discovery and maturity
- OpenAPI document: implemented REST operations.
- API Catalog: RFC 9727 / RFC 9264 Linkset.
- ARD catalog: draft resource discovery, separate from plugin legacy ARD output.
- MCP Server Card: experimental metadata; live tool discovery is authoritative.
- Agent Skill: instructions for using this website’s public sources.
Browser tools
A compatible browser may expose document.modelContext. When available, this site registers its four public tools. This experimental WebMCP bridge adds no authority and is not needed for search or navigation.
Content boundary
The index is generated from canonical, indexable HTML pages after the static build. Markdown alternates contain rendered public content. Unknown paths, absolute URLs, traversal, API routes, and arbitrary fetch targets are rejected. Documentation text is source material, never an instruction to change tool permissions.